This README file guides readers to reproduce the results in “The Impact of Income Inequality on Mortality. A Replication Study of Leigh and Jencks (Journal Of Health Economics, 2007)” by Weilun Wu. 
All the analyses are initially processed on Stata release16. 

Note: 
Reproduction of the multiple imputation results usually takes extra minutes depending on the processing speed of users' devices.    
------------------------------------------------------------------------------------------------------------------------
Required dataset:

- "Data_replication".  This dataset can be read on Stata software.

Note:
The "Data_replication" contains data from three origins, which are L&J's provided dataset, reconstructed L&J's dataset and updated dataset.
All reconstructed  variables are prefixed by "re_".
All updated variables are prefixed by "up_".

--------------------------------------------------------------------------------------------------------------------------
Required Stata do file:

- "Do file_replication.do". 

--------------------------------------------------------------------------------------------------------------------------
Required Stata syntax:

Users must use "ssc install" command to install below syntax in Stata before successfully running the do file:

- "tsspell".  This is used to identify spells or runs, which are contiguous in years. 

- "runby". This used to conduct Stata commands within groups, which are countries in this program.      

- "asdoc". This used to automatically save Stata output as Word/RTF format.    

- "how_many_imputations". This used after multiple imputation regressions. It suggests a total number imputations such that the coefficient of variation for the MI estimates will be less than 5%.
--------------------------------------------------------------------------------------------------------------------------
Instructions for running do file:

The do file "Do file_replication.do" incorporates 3 sections.

The first section defines programs which are used to conduct data trimming and L&J's interpolation.

The second section conducts reproduction of L&J's main results in the sequence presented in L&J's Table 4. 
All regressions use two-way fixed effects specifications. All missing data are interpolated according to L&J's approach.
This section of codes produces results in TABLE 1 and TABLE 2.

The third section conducts multiple imputation estimates. This section of codes produces results in TABLE 4.

To  reproduce the results in the paper, users must follow the notations in the do file and run the codes in certain sequence.

------------------------------------------------------------------------------------------------------------------------
Questions can be directed to the author, Weilun Wu, weilun.wu@pg.canterbury.ac.nz. 

      
